key: cord-0315827-o4vvlmr4
authors: Wulf Hanson, S.; Abbafati, C.; Aerts, J. G.; Al-Aly, Z.; Ashbaugh, C.; Ballouz, T.; Blyuss, O.; Bobkova, P.; Bonsel, G.; Borzakova, S.; Buonsenso, D.; Butnaru, D.; Carter, A.; Chu, H.; De Rose, C.; Diab, M. M.; Ekbom, E.; El Tantawi, M.; Fomin, V.; Frithiof, R.; Gamirova, A.; Glybochko, P. V.; Haagsma, J. A.; Haghjooy Javanmard, S.; Hamilton, E. B.; Harris, G.; Heijenbrok-Kal, M. H.; Helbok, R.; Hellemons, M. E.; Hillus, D.; Huijts, S. M.; Hultstrom, M.; Jassat, W.; Kurth, F.; Larsson, I.-M.; Lipcsey, M.; Liu, C.; Loflin, C. D.; Malinovschi, A.; Mao, W.; Mazankova, L.; McCulloch, D.; Menges, D
title: A global systematic analysis of the occurrence, severity, and recovery pattern of long COVID in 2020 and 2021
date: 2022-05-27
journal: nan
DOI: 10.1101/2022.05.26.22275532
sha: 562b71260f906a7801a4781d03e1c7d694a5948a
doc_id: 315827
cord_uid: o4vvlmr4

Importance: While much of the attention on the COVID-19 pandemic was directed at the daily counts of cases and those with serious disease overwhelming health services, increasingly, reports have appeared of people who experience debilitating symptoms after the initial infection. This is popularly known as long COVID. Objective: To estimate by country and territory of the number of patients affected by long COVID in 2020 and 2021, the severity of their symptoms and expected pattern of recovery Design: We jointly analyzed ten ongoing cohort studies in ten countries for the occurrence of three major symptom clusters of long COVID among representative COVID cases. The defining symptoms of the three clusters (fatigue, cognitive problems, and shortness of breath) are explicitly mentioned in the WHO clinical case definition. For incidence of long COVID, we adopted the minimum duration after infection of three months from the WHO case definition. We pooled data from the contributing studies, two large medical record databases in the United States, and findings from 44 published studies using a Bayesian meta-regression tool. We separately estimated occurrence and pattern of recovery in patients with milder acute infections and those hospitalized. We estimated the incidence and prevalence of long COVID globally and by country in 2020 and 2021 as well as the severity-weighted prevalence using disability weights from the Global Burden of Disease study. Results: Analyses are based on detailed information for 1906 community infections and 10526 hospitalized patients from the ten collaborating cohorts, three of which included children. We added published data on 37262 community infections and 9540 hospitalized patients as well as ICD-coded medical record data concerning 1.3 million infections. Globally, in 2020 and 2021, 144.7 million (95% uncertainty interval [UI] 54.8-312.9) people suffered from any of the three symptom clusters of long COVID. This corresponds to 3.69% (1.38-7.96) of all infections. The fatigue, respiratory, and cognitive clusters occurred in 51.0% (16.9-92.4), 60.4% (18.9-89.1), and 35.4% (9.4-75.1) of long COVID cases, respectively. Those with milder acute COVID-19 cases had a quicker estimated recovery (median duration 3.99 months [IQR 3.84-4.20]) than those admitted for the acute infection (median duration 8.84 months [IQR 8.10-9.78]). At twelve months, 15.1% (10.3-21.1) continued to experience long COVID symptoms. Conclusions and relevance: The occurrence of debilitating ongoing symptoms of COVID-19 is common. Knowing how many people are affected, and for how long, is important to plan for rehabilitative services and support to return to social activities, places of learning, and the workplace when symptoms start to wane.

Much of the attention of disease surveillance of the coronavirus disease 2019 pandemic has 202 concentrated on the number of infections, the large number of cases requiring hospital care for severe 203 infection, and those who have died from the disease. Less attention has been given to the quantification 204 of those who continue to experience symptoms past the acute infection stage. Terms such as long 205 COVID, COVID long haulers, brain fog, post-COVID-19 condition, or post-acute sequelae of have been used to describe a diverse array of ongoing symptoms. In October 2021, the World Health 207

Organization released a clinical case definition for post-COVID-19 condition as symptoms that are 208 present at three months after SARS-CoV-2 infection with a minimum duration of 2 months and cannot 209 be explained by an alternative diagnosis. 1-6 We will use the term long COVID in this paper. 210

Post-acute infection fatigue syndromes have been described for other viruses and bacteria. 7-10 An 211 ongoing low-grade inflammation has been postulated to cause these symptoms, but the pathology 212 remains largely unknown and treatments are based on symptom relief. 2, 11, 12 The impact on affected 213 individuals is substantial, and special clinics dealing with patients of long COVID have arisen to respond 214 to an increasing need for supportive and rehabilitative care. 13-16 215 A meta-analysis of 45 follow-up studies of COVID patients, of which only three had a follow-up time 216 greater than three months, found 84 long-term effects of COVID-19, with shortness of breath, fatigue, 217 and sleep disorders or insomnia as the most common symptoms. 17 Studies have most frequently just 218 reported on individual symptoms or counts of symptoms but less on severity, overlap of symptoms and 219 pattern of recovery. 1,18-22 220 In this paper, we have collated the information on long COVID into three common clusters of symptoms 221 largely based on joint analyses with custodians of follow-up studies after COVID-19 diagnosis in ten 222 countries, supplemented by published data and medical record databases. From this pooled information 223 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 Input data 249

From the long list of persisting symptoms reported by a proportion of COVID-19 cases after the acute 250 phase, we selected three major symptom clusters based on frequency and the ability to quantify their 251 relative severity using health state descriptions and corresponding disability weights (DWs) from the 252 Global Burden of Disease (GBD) study. GBD uses 236 health states and DWs to quantify the non-fatal 253 consequences of diseases and injuries. 28 Table 1 shows the health states, lay descriptions, and disability 254 weights used for long COVID. The three symptom clusters were a) fatigue with bodily pains and/or 255 symptoms of depression or anxiety; b) cognitive problems such as forgetfulness or difficulty in 256 concentrating, commonly referred to as "brain fog"; and c) ongoing respiratory problems with shortness 257 of breath and persistent cough as the main symptoms. We decided to distinguish between two severity 258 levels for cognitive problems and three levels of severity for the ongoing respiratory symptoms. From 259 here on, we refer to the "fatigue", "respiratory", and "cognitive" clusters. 260

We conducted a systematic review of published papers on the long-term consequences of COVID-19 but 261 found that no published study provided enough detail for our quantification purposes. From 7362 262 unique search hits and additions screened from a living systematic review 29 , we included 46 published 263 articles from 44 studies that contain follow-up data of at least one defining symptom included in our 264 defined symptom clusters (Supplementary Appendix Data Inputs). We used the Preferred Reporting 265

Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines; 30 our PRISMA checklist is in the 266 appendix (eSection 5), and the search protocol was registered with the International Prospective 267

Register of Systematic Reviews (PROSPERO, CRD42020210101). 31 268 Instead of relying on published reports only, we contacted study authors of published studies and 269 ongoing COVID-19 follow-up studies that were registered at the ISRCTN registry. 32 From 23 positive 270 responses of 42 study authors contacted, ten were able to share symptom cluster data in time for 271 inclusion in our study (Table 2 ). With researchers from the ten follow-up studies, we developed 272 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 algorithms to define the three symptom clusters by severity level by choosing symptom questions and 273 measures employed in each study that would most closely match the wording of the lay descriptions 274 that were presented to respondents of the GBD DW surveys (Table 1 ). In the Supplement eSection 5, we 275 present details of the algorithms for each of the included studies. We utilized the cohort data with 276 explicit questions comparing current symptoms to those pre-COVID to adjust the remainder of the 277 cohort data lacking pre-COVID comparisons ( In addition, we received analyses from collaborators at the Veterans Affairs Health Administration and 279 from PRA Health Sciences, a data collection of private health insurance plans, based on ICD codes for the 280 primary symptoms belonging to the three symptom clusters of interest among COVID patients 281 compared to matched non-COVID patients (Supplementary Appendix Data Inputs). 33-35 ICD codes are 282 provided in the Supplement eTable 4. 283

We first undertook a meta-regression of studies with multiple follow-up measurements to determine 285 the recovery pattern of symptoms. Given the relative scarcity of data, we assumed a similar pattern of 286 recovery for all three symptom clusters. We used separate models for community cases and hospitalized 287 cases using a Bayesian meta-regression tool, MR-BRT (meta-regression-Bayesian, regularized, 288 trimmed), to pool the logit-transformed proportions of cases with any of the three symptom clusters by 289 follow-up time since the end of the acute episode (eFigure 10). 36,37 For community cases, we used data 290 from the Zürich and Faroe studies supplemented with data derived from three published studies. 18-20,38-291 40 With a dummy in the meta-regression, we adjusted the Cirulli, United Kingdom COVID Symptom 292 Study (CSS), and United Kingdom COVID-19 Infection Survey (CIS) studies as proportions were reported 293 for aggregates of many individual symptoms, rather than only the symptom clusters of our interest. 18-294 20,40 For hospitalized cases, we used data from the COVID-19 Follow-up care paths and Long-term 295

Outcomes Within the Dutch health care system (CO-FLOW), Sechenov StopCOVID, PronMed ICU, and 296 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 13 the Zürich SARS-CoV-2 Cohort studies supplemented with data derived from two published studies in 297 Switzerland and Spain, both adjusted as in the community cases duration model. 38, 41 The longest follow-298 up from these studies was 12 months in the Zürich, Co-FLOW and Sechenov studies. In both models, an 299 exponential decline was assumed in the proportion of cases affected by long COVID. The coefficients on 300 the rate of decline in these initial models were then entered as priors into the models that used all 301 available follow-up data, described below. 302

Next, we used all the data in models of long COVID in community and hospitalized cases separately. We 303 ran separate models for each of the three clusters and the overlap between clusters and adjusted their 304 outputs proportionally to sum to the values of the models for any of the three clusters of long COVID 305 (eFigures 11-15). We had too few data points to run separate models for ICU-admitted cases; in the 306 hospital models for each symptom cluster, we used a dummy variable for those admitted to ICU in order 307 to predict their proportions, with the coefficient informed by the observed relationship between ICU 308 and non-ICU hospitalized data. 33, 35 We also included variables for sex, whether the data were from an 309 administrative source, and indicator variables for individual symptoms reported in the published articles 310 (cognitive dysfunction, shortness of breath, fatigue). 311

To estimate the severity distributions, we pooled data from cohorts that had enough detail to determine 312 the two levels of severity of cognitive symptoms and the three levels of severity of ongoing respiratory 313 problems using a random effects meta-analysis with a fixed effect on hospitalized data (eFigures 16-17). 314

Incidence, prevalence, and severity-weighted prevalence of long COVID 315

To estimate the incidence of long COVID, we first subtracted deceased patients from the incidence 316 numbers of symptomatic COVID infection and then multiplied these surviving patients by the estimated 317

proportions of cases with each symptom cluster at 3 months. 42,43(p19),44-46 Daily incident cases of long 318 COVID at three months post-infection were multiplied by the average duration, estimated separately for 319 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 14 community cases and hospitalized cases. We then summed the prevalent days of long COVID for each of 320 the symptom clusters and their overlap by level of severity where applicable across the years 2020 and 321 2021. Each of these was multiplied by the corresponding DW to get severity-weighted prevalence, 322 equivalent to the GBD metric of years lived with disability (YLDs). For overlapping clusters, we assumed a 323 multiplicative function to constrain the combined DW to a value between zero and one. 47 324

We present uncertainty intervals (UIs) for all estimates based on the 25 th and 975 th values of the ordered 325 1000 draws of the posterior distributions. 326 327 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 

Globally within 2020 and 2021, of 3.92 billion (95% UI 3.77-4.05) infections with SARS-CoV-2 through 329 the end of 2021, 3.7% (1.4-8.0) or 144.7 million (54.7-312.6) persons developed long COVID defined as 330 experiencing one or more of the three symptom clusters three months after infection (Table 3) infections is much higher than reported as diagnosed cases because excess deaths, infection-to-death 339 ratios and seroprevalence surveys suggest many more cases must have occurred. We estimated a 340 median duration of long COVID of 3.99 months in community infections, while 341 hospitalized cases were estimated to experience a longer median duration of 8.84 months (IQR 8.10-342 9.78) (eFigure 10). The global prevalence of long COVID in 2020-2021 was 5.11 million (2.31-8.72) cases 343 among more severe, hospitalized patients and 31.4 million (10.2-73.5) cases among those who had 344 milder infections. 345

The fatigue, respiratory, and cognitive clusters occurred in 51.0% (16.9-92.4), 60.4% (18.9-89.1), and 346 35.4% (9.4-75.1) cases of long COVID, respectively. In 38.4% (7.94-96.0) of long COVID cases, two or all 347 three of the clusters overlapped ( Figure 1 ). 348

Globally, among prevalent long COVID cases, 63.2% (59.7-66.3) were female. The risk of long COVID at 3 349 months follow-up under the age of 20 was lower than in adults in milder community infections, 2.73% 350 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 (0.808-6.65) in children versus 4.76% (1.53-11.3) in adult males and 9.88% (3.38-21.2) in adult females 351 (eTable 15a). The peak ages of long COVID cases were between 20 and 29 ( Figure 2) . 352

The average level of disability among long COVID cases, estimated as the ratio of overall long COVID 353 severity-weighted prevalence to prevalence, was 0.231 (0.134-0.370)-equivalent to the GBD DWs for 354 severe neck pain, Crohn's disease, or long-term consequences of moderately severe traumatic brain 355 injury. 28 356

The age and geographical pattern of incidence and prevalence of long COVID closely follows that of 357 SARS-CoV-2 infections, as we assumed the same risk among survivors of acute infection, severity 358 distributions, and duration in all locations. The counts of incident and prevalent cases of long COVID by 359 country are provided in Table 4 . 360

Among COVID patients who develop long COVID in 2020 and 2021, 15.1% (10.3-21.1) continued to have 361 persistent symptoms at 12 months after COVID infection, or 21.0 million (9.19-41.7) people. In the 362 United States, the Social Security Administration is mandated to financially assist those who are still 363 unable to work after 12 months, corresponding to 946,000 (365,000-2,130,000) people aged 20-64 in 364 the US in 2021 and 2022. 365 366 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 Discussion 367 A substantial proportion of COVID-19 patients do not recover after the initial infection. We estimated 368 that 144.7 million (54.7-312.6) cases globally in 2020 and 2021 suffered from one or more of three 369 common symptom clusters of long COVID. The risk of long COVID is greater in females and in those with 370 more severe initial infection. The peak ages of those experiencing long COVID were between 20 and 29 371 years. This pattern by age and sex is distinct from that of severe acute infection, which affects more 372 males and increases with age. 48 From seventeen follow-up studies that included children, we also know 373 that long COVID affects a lower but substantial number of children, while severe acute infection is very 374 uncommon at younger ages. 49-51 These differences suggests that the underlying mechanism of long 375 COVID may be different from that of the severity of acute infection. 376

A prolonged state of low-grade infection with a hyperimmune response, coagulation/vasculopathy, 377 endocrine and autonomic dysregulation, and a maladaptation of the angiotensin converting enzyme-2 378 (ACE-2) pathway have been postulated as the underlying pathophysiology of long COVID. 52 379

Deconditioning due to prolonged immobilization during hospitalization may compound these 380 problems. 53 Direct tissue damage due to COVID-19 has been demonstrated in many parts of the body, 381 including the lung, heart, kidney, and brain. [54] [55] [56] [57] [58] [59] Due to the large reserves in capacity of most body 382 organs, tissue damage does not immediately lead to symptomatic disease. It may, however, become 383 apparent over time that COVID-19 will contribute to an earlier onset and greater occurrence of long-384 term symptomatic major organ disease with increasing age or if these organs become diseased by other 385 mechanisms. The rate of recovery from long COVID, moreover, suggests that less permanent factors 386 may underlie these debilitating symptoms. 387

We have adopted the WHO case definition which stipulates a minimum period of three months after 388 infection before calling ongoing symptoms long COVID or post-COVID-19 condition. Others have 389 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 suggested a threshold of three weeks to define a case of long COVID, arguing that no competent virus 390 has been replicated beyond three weeks of infection, but a period of up to 12 weeks has been suggested 391 to define the start of long COVID. 12, 52, 60 This analysis accounts for COVID infections through the end of 392 2021 and therefore does not cover the omicron wave, and it is currently unclear what the risk of long 393 COVID is after infection with omicron. The large proportion of asymptomatic infections with omicron 394 and the fact that if symptoms arise they mostly affect the upper airways suggests that the risk of long 395 COVID will be much smaller. 396

The recovery pattern among community cases for the three symptom clusters quantified suggests that 397 the majority of cases resolve, a sign of hope for those experiencing these debilitating symptoms. It is not 398 yet clear if there is a smaller proportion of patients, especially among those with more severe acute 399 episodes, who develop a more chronic course of long COVID. Given that the longest follow-up time 400 among the studies we examined was 12 months, the true long-term pattern of recovery will only be 401 revealed as studies conduct longer follow-up periods. The time-limited course of long COVID in most 402 cases has led to the advice to provide rehabilitative support in the community, with specialist 403 rehabilitation services required only for those with protracted and more severe problems, particularly 404 when compounded by post-intensive care syndrome. 12,61 Important to patients is that they feel empathy 405 and recognition from health-care workers even if they can only provide symptomatic and supportive 406 care. 62 Quantifying the number of incident and prevalent cases of long COVID will help policy makers 407 ensure adequate access to services to guide patients towards recovery, return to the workplace or 408 schooling, and restoration of their mental health and social life. The attention given to long COVID may 409 also provide greater recognition to patients who suffer from the longer-term consequences of other 410 infectious diseases and who may have received less attention from health services. The large number of 411

people affected by long COVID should also create new opportunities to unravel phenotypical and 412 genotypical characteristics, with an aim to find new treatments and predictors of post-acute disease 413 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22275532 doi: medRxiv preprint syndromes including those known to occur after other infectious disease and intensive care for other 414 critical illness. 415

The main strength of this study is the willingness of researchers from ten follow-up studies to share data 416 and analyses with consistent approaches to deal with the diverse study methods and instruments. This 417 collaborative effort also allowed us to go beyond the reporting of individual symptoms or counts of 418 symptoms reported in the literature. With access to individual patient data, we were able to define 419 clusters of symptoms that frequently occur together and to quantify the overlap among symptom 420 clusters. Importantly, we were able to correct for over-reporting from studies that did not have a 421 comparison with previous health status, leveraging information from the cohort studies that explicitly 422 asked respondents to recall their pre-COVID health status or existence of symptoms. In addition, the 423 very large health insurance databases from the USA allowed us to identify controls matched on 424 demographic and disease characteristics and thus correct for the occurrence of these symptoms 425 unrelated to SARS-CoV-2 infection. This may in part explain why our estimates of long COVID are lower 426 than often reported in the literature. Direct comparisons are not easy, as we have defined clusters of 427 symptoms that are not reported by others. However, we think this is a strength of this analysis in 428 comparison to studies reporting individual symptoms or counts of symptoms. 429

There are also important limitations to our analysis. First, the uncertainty intervals around the estimates 430 are wide, reflecting as yet limited and heterogeneous data. Second, we had to derive separate 431 algorithms for each contributing study to achieve consistency in case definitions of the three chosen 432 symptom clusters. Efforts to achieve standardization of questions and instruments for studies of long 433 COVID are underway. 5,63 This would make pooling estimates among studies less prone to measurement 434 bias. Third, we assumed that long COVID follows a similar course in all countries and territories. We used 435 data from western European countries, Iran, Russia, India, China, South Africa, Turkey, Saudi Arabia, 436

Israel, Australia, and the USA. Additional reports from Brazil and Bangladesh suggest that long COVID 437 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22275532 doi: medRxiv preprint 20 similarly affects other parts of the world. 21,22 As more information becomes available, we can explore 438 whether there is geographical variation in the occurrence or severity of long COVID. We also note that 439 the duration estimates relied on studies from high income countries only. With repeated follow-up 440 being planned in many of the studies and with new studies appearing, it will become clearer over time 441 how generalizable our findings on duration are. Fourth, apart from symptoms and symptom clusters, 442 new diseases have been reported to occur more frequently in patients after COVID-19 diagnosis, 443 including cardiovascular complications like myocarditis, acute myocardial infarction, and thrombo-444 embolic events as well as kidney, liver, gastrointestinal, endocrine, and skin disorders. [64] [65] [66] The data 445 sources to quantify these COVID-19-related changes may not yet be sufficient due to lags in reporting of 446 clinical informatics data, disease registries, or surveys, which form the basis of estimation for such 447 diseases. Fifth, with limited follow-up time available, the pattern of recovery cannot yet be fully 448 described. Importantly, longer follow-up can reveal if there is a subset of cases that go on to have a 449 protracted course of long COVID and need longer care. Sixth, we made the assumption that long COVID 450 only affects those with a symptomatic course of the initial infection. The participating cohorts included 451 few asymptomatic cases: the Faroe Islands, Zurich SARS-CoV-2 Cohort, HAARVI, Rome ISARIC pediatrics 452 and adults cohorts observed 22, 182, 9, 27, and 26 asymptomatic COVID cases, respectively. Long 453 COVID was not identified among asymptomatic cases that were followed in HAARVI and Rome ISARIC 454 cohorts. In the Faroe Islands and Zurich SARS-CoV-2 cohorts, three and five of their asymptomatic cases, 455 respectively, developed at least one long COVID symptom cluster at follow-up. The total number of 456 asymptomatic cases followed in these studies is very low and we chose to be cautious and exclude them 457 from our calculations. In a review of medical records in the University of California COvid Research Data 458 Set (UC CORDS), 32% of those with long COVID symptoms at two months after a positive PCR test 459

reported no symptoms at testing, but it is not clear how many of these developed acute symptoms after 460 testing. 60 Seventh, we chose three commonly reported symptom clusters but have not quantified other 461 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22275532 doi: medRxiv preprint 21 common symptoms. The main symptoms of our three symptom clusters are those that reached the 462 highest degree of consensus in the Delphi process WHO used to create a clinical case definition for post 463 In the most complete cohort, the Sechenov StopCOVID cohort, we had information 464 on a wide range of symptoms and general health status with explicit comparison with the pre-COVID-19 465 status (eSection 5). Among 1309 respondents with PCR-confirmed COVID-19 needing hospitalization, 466 136 qualified for at least one of our three symptom clusters of long COVID at six months follow-up. 467

Another 62 respondents reported not having fully recovered. Of these, 48 reported at least one 468 symptom of our three symptom clusters but had failed to meet all criteria by reporting either no or 469 slight problems with usual activities or no worsening of this item compared to before COVID-19. Other 470 more common symptoms that were reported by this group included problems with vision, sleep 471 problems, loss of smell, palpitations, and hair loss. Quantifying vision loss requires measurement of 472 visual acuity, which is not measured in long COVID studies. There are no DWs for loss of smell, hair loss, 473 or palpitations. While there is a disability weight for insomnia, it has not been used in any GBD study as 474 sleep disorders are not (yet) included in the GBD cause list. Estimates therefore do not reflect the 475 burden of the full range of long COVID outcomes. 476

We have quantified the frequency at which common symptom clusters of long COVID have occurred 478 across the world and made an estimate of their severity and expected duration. Many countries and 479 territories have already responded by setting up specialized treatment centers for those affected. 67,68 480

Understanding the magnitude of the problem will help other countries and territories to respond 481 likewise. Early studies indicate that for most patients with long COVID, there is hope for recovery, but 482 time will tell if all patients recover. The attention given to long COVID during this pandemic should 483 trigger research into the underlying pathology and potential treatment or prevention, the long-term 484 trajectory of long COVID, the potential transition from long COVID into chronic fatigue syndrome, the 485 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22275532 doi: medRxiv preprint 22 level of protection from vaccination and the risk of long COVID following more recent omicron variants. 486

Such research may also benefit those who experience similar outcomes following a range of other 487 infectious diseases, an issue that has not received much attention from clinical and global heath 488 communities. 489

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The funders of the study had no role in study design, data collection, data analysis, data interpretation, 567

or the writing of the report. Members of the core research team for this topic area had full access to the 568 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22275532 doi: medRxiv preprint 25 underlying data used to generate estimates presented in this paper. All other authors had access and 569 reviewed estimates as part of the research evaluation process, which includes additional stages of 570 formal review. 571 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. (7):1017-1032. doi:10.1038/s41591-020-0968-3 749 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 Main text figures and tables is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) Figure 2 . Global incident cases, prevalent cases, and severity-weighted prevalence of long COVID by age, sex, symptom cluster, and overlap of symptom clusters in 2020 and 2021 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 

NICE guideline on long COVID

Post-COVID-19 Symptom 609 Burden: What is Long-COVID and How Should We Manage It? Lung

Developing services for long COVID: lessons from a study 612 of wounded healers

Post COVID-19 Syndrome (Long Haul Syndrome): 614 Description of a Multidisciplinary Clinic at the Mayo Clinic and Characteristics of the Initial Patient 615

Assessment of the Frequency and Variety of Persistent 617 Symptoms Among Patients With COVID-19: A Systematic Review

Long-term COVID-19 symptoms in a large unselected 620 population. medRxiv

Updated estimates of the prevalence of long COVID symptoms -623 Office for National Statistics

Attributes and predictors of long COVID

Post-Acute COVID Syndrome, the Aftermath of 629 Mild to Severe COVID-19 in Brazilian Patients. medRxiv

Post-COVID-19 syndrome among symptomatic COVID-632 19 patients: A prospective cohort study in a tertiary care center of Bangladesh

Guidelines for Accurate and Transparent Health Estimates 635 Reporting: the GATHER statement

Severe acute kidney injury 750 associated with progression of chronic kidney disease after critical COVID-19

Clinics are springing up around the country for what some call a potential second 753 pandemic: Long Covid

Brain fog, fatigue, breathlessness. Rehab centers set up 756 across Europe to treat long-term effects of coronavirus

Neurological outcome and quality of life 3 months after 759 COVID-19: A prospective observational cohort study

Isfahan COVID Cohort" 762 Study: Rationale, Methodology and Initial results

Studying the pathophysiology of coronavirus disease 764 2019: a protocol for the Berlin prospective COVID-19 patient cohort (Pa-COVID-19)

Clinical and virological characteristics of hospitalised 768 COVID-19 patients in a German tertiary care centre during the first wave of the SARS-CoV-2 769 pandemic: a prospective observational study

PRONMED Uppsala COVID-19 ICU Biobank

Impaired diffusing capacity for carbon monoxide is common in 774 critically ill Covid-19 patients at four months post-discharge

Asthma is associated with Long Covid in previously 777 hospitalised adults: StopCOVID cohort study

Post-COVID-19 Syndrome: Nine Months after SARS-CoV-2 785 Infection in a Cohort of 354 Patients: Data from the First Wave of

Patient outcomes after hospitalisation with COVID-19 and 788 implications for follow-up: results from a prospective UK cohort

Long COVID in children and adolescents

Long COVID 1 year after hospitalisation for COVID-19: a 793 prospective bicentric cohort study

Respiratory and Psychophysical Sequelae Among Patients 796 With COVID-19 Four Months After Hospital Discharge

Long COVID symptoms in SARS-CoV-2-positive adolescents 799 and matched controls (LongCOVIDKidsDK): a national, cross-sectional study. Lancet Child Adolesc 800 Health

Gemelli Against COVID-19 Post-Acute Care Study Group. Persistent 802 Symptoms in Patients After Acute COVID-19

Follow-up of adults with noncritical COVID-805 19 two months after symptom onset

Clinical predictors of long COVID-19 and phenotypes of 808 mild COVID-19 at a tertiary care centre in India

Sixty-Day Outcomes Among Patients 811 Hospitalized With COVID-19

Long COVID and the mental and physical health of 813 children and young people: national matched cohort study protocol (the CLoCk study)

One-Year Follow-up Cohort Study. Open Forum 817 Infect Dis

Russian Federation 1130

122-738) the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted

. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022